This was discussed previously : http://lists.freedesktop.org/archives/dri-devel/2012-January/018463.html
The plat de resistance is in patch4, the first 3 are simplification that i believe can go as is.
Idea is that driver needs to know when ever a bo move happen and instead of having 3 set of callback to allow driver to do things btw common ttm move handling. Move the ttm move handling into each driver and let driver customize it at appropriate point.
I intentionaly didn't not fully take advantage of that in neither radeon or nouveau. ie i tried to keep the current code as much as possible so one can see that i don't disrupt old behavior. There is opportunity to simplify the move_notify driver function i believe.
I tested it on radeon and nouveau. This time i limited the 2G vram nvidia gpu i have to behave like a 64M vram or 96M vram so i actually see the vram eviction path being exercice while testing (this is why in my last ttm patch set i missed the nouveau regression, again sorry about that). So i am more confident that with this one i didn't regressed anything.
Thomas i tried to be carefull with vmwgfx but i didn't tested it. Only built tested.
Cheers, Jerome